SAFE: Secure Authentication with Face and Eyes Arman Boehm, Dongqu Chen, Mario Frank, Ling Huang, Cynthia Kuo, Tihomir Lolic, Ivan Martinovic, Dawn Song Face authentication is commonly offered as an alternative to passwords for device unlock. However, commercially available face authentication systems are vulnerable to simple spoofing attacks. We demonstrate the impact of image quality on spoofing, using low resolution photos representative of those commonly posted online. We also show that videos, slideshows of images at different angles, and crude 3D avatars are effective. To defend against these vulnerabilities, we propose a face authentication system that includes a secrecy challenge. We present SAFE (Secure Authentication with Face and Eyes), an improved face authentication method that uses a commodity gaze tracker to input a secret. During authentication, the user must not only show her face but also gaze at a secret icon that moves across the screen. Using a novel method for estimating the noise level in the gaze tracking data, SAFE adapts the system's parameters to enable secure, hands-free authentication.